InsilicoΣ
Drug Discovery, Cheminformatics & Bioinformatics
About Our Team Publications Contact Us Login Register

About InsilicoΣ

An academic research lab for computational drug discovery, cheminformatics, and bioinformatics, bridging the gap between scientific research and practical tools that accelerate real-world discovery.

Our Mission

InsilicoΣ is an academic research laboratory focused on computational drug discovery, cheminformatics, and bioinformatics. We believe that every computational method, every predictive model, and every validated algorithm deserves a second life, not just as a paper, but as a working tool that any researcher can use from day one.

"We don't just publish science, we deploy it. Every model we develop is built to be used, not just cited." ~ Salah A.

Our mission is to transform peer-reviewed computational research into accessible, production-ready software tools that researchers around the world can use to accelerate drug discovery, cheminformatics, bioinformatics, material design, and molecular engineering, without needing to reproduce complex pipelines themselves.

Σ What Does Σ (Sigma) Mean?

The Greek letter Σ (Sigma) is the universal symbol for summation in mathematics and science. It represents the idea of bringing many parts together into one meaningful whole. This is exactly what our platform does.

Every tool on InsilicoΣ is built on a philosophy of aggregation and consensus. Rather than relying on a single model or a single data source, we combine multiple prediction engines, multiple tiers of computation, and multiple lines of evidence to produce results that are more reliable than any single method alone.

"Σ = summation. We aggregate engines, tiers, and evidence. One prediction is an estimate; a consensus is knowledge."

This philosophy runs through every tool we build:

ADMET-Σ

Aggregates 3 prediction tiers plus 10 enhanced features including consensus scoring, prediction reliability indexing, and clinical developability assessment.

POLY-X

Fuses Group Contribution theory, machine learning (Random Forest + Gradient Boosting), and transformer-based polyBERT embeddings into a unified prediction.

CRAFT

Creates a 256-bit biological-context fingerprint from four hierarchical modules, then fuses it with ECFP4 molecular fingerprints into a 2304-bit "drug-in-context" vector.

QSAR-X

Supports 2D, 3D, and 4D descriptor spaces with ensemble model validation, applicability domain analysis, and SHAP-based explainability converging into trustworthy predictions.

RNA-Σ

Orchestrates 4 RNA therapeutic modalities (siRNA, sgRNA, ASO, mRNA) through a unified design pipeline with integrated SafeRNA immunogenicity profiling (TLR + IFN + off-target scoring).

RoB-Σ

Aggregates domain-level judgments across 7 Risk of Bias instruments using instrument-specific algorithms (worst-case for RoB 2, star-counting for NOS, critical-flaw for AMSTAR 2) into unified study-level risk assessments.

GRADE-Σ

Synthesizes rule-based AI assessments across 8 GRADE domains (5 downgrade + 3 upgrade) with deterministic imprecision, inconsistency, and publication bias heuristics into a unified certainty-of-evidence rating.

Meta-Cleaner

Standardizes heterogeneous study data before pooling — converting units across 13 analytes and deriving SD from any reported dispersion measure (SEM, 95% CI, IQR, range) — so downstream meta-analyses begin from a consistent, harmonized dataset.

Why InsilicoΣ?

Computational research in drug discovery, cheminformatics, and bioinformatics generates thousands of predictive models and methods every year. Most of these remain as static figures and tables in journal articles, inaccessible to the broader research community. Reproducing published methods often requires months of effort, specialized infrastructure, and deep domain expertise.

InsilicoΣ was created to solve this problem. We take the methods we develop and publish, and turn them into interactive web-based tools that are ready to use out of the box. No installation, no coding required, just science, applied.

Our Approach

  • Research-Driven Development, Every tool on InsilicoΣ is grounded in peer-reviewed, validated science. We develop the methodology first, publish it, and then build the tool.
  • Accessibility Over Complexity, Sophisticated computational methods should not require a PhD to operate. Our platform is designed so that experimentalists, medicinal chemists, and students can benefit from advanced modeling without writing a single line of code.
  • End-to-End Workflows, From molecular input to actionable predictions, each tool covers the full pipeline: data preparation, descriptor calculation, model training, validation, and prediction, all in one place.
  • Open Science, Applied, We don't gatekeep methods behind paywalls or complex installations. InsilicoΣ makes cutting-edge computational science available to anyone with a browser.
  • Continuous Evolution, As our research advances, so do the tools. New publications feed directly into platform updates, keeping every tool current with the latest science.

The Platform

InsilicoΣ hosts a growing suite of computational tools spanning drug discovery, cheminformatics, bioinformatics, and molecular design. The platform is organized into two categories:

Novel Tools — Developed by InsilicoΣ

Original tools built from our own peer-reviewed research and published methodologies.

QSAR-X

Quantitative Structure-Activity Relationship

Build, validate, and deploy QSAR models with full support for 2D, 3D, 4D, and interaction-resolved descriptors. From data upload to prediction, one integrated workflow.

ADMET-Σ

ADMET Property Prediction

Predict absorption, distribution, metabolism, excretion, and toxicity properties for drug candidates using validated multi-endpoint prediction models.

APPAS

Automated Polymer Property Analysis

Predict physicochemical properties of copolymers from monomer structures and sequences. The first web-based tool for automated co-polymer property prediction.

NEXUS

Network Pharmacology Analysis

Map compound-target-disease interaction networks. Identify multi-target mechanisms of action through pathway enrichment and network topology analysis.

POLY-X

Polymer Property Prediction

Predict thermal and physical properties of polymers using a 3-tier system: group contribution, machine learning, and polyBERT transformer embeddings with consensus scoring.

CRAFT

Bio-Context Target Fingerprint

Generate 256-bit biological-context fingerprints for drug targets, encoding membrane topology, signal transduction, binding pocket geometry, and endogenous ligand properties.

RNA-Σ

Unified RNA Therapeutic Design

Design siRNA, sgRNA (CRISPR), ASO/GapmeR, and mRNA with integrated SafeRNA immunogenicity profiling, target accessibility scoring, and off-target seed analysis.

RoB-Σ

Risk of Bias Assessment

Guided risk of bias assessment using 7 validated instruments (RoB 2, ROBINS-I, NOS, QUADAS-2, QUIPS, AMSTAR 2) with automated judgment algorithms and publication-ready figures.

GRADE-Σ

AI-Augmented GRADE Evidence Evaluation

Full GRADE framework with rule-based AI suggestions for 8 domains, OIS calculation, I² interpretation, Summary of Findings tables, and Cochrane-style Word/PDF export.

Meta-Cleaner

Meta-Analysis Data Standardization

Convert units (13 analytes, bidirectional) and derive SD from SEM, 95% CI, IQR, or range using Wan 2014 methods. Outputs a clean, color-coded table with CSV/Excel export ready for the next meta-analysis step.

Adapted Tools — Open-Source Integrations

Leading open-source research tools adapted with user-friendly web interfaces by InsilicoΣ. Full attribution to original authors is maintained throughout the platform.

REINVENT4

AI-Driven Molecular Generation

Generate novel drug-like molecules using deep generative models. Design compounds optimized for target activity, selectivity, and drug-likeness through reinforcement learning.

DecompDiff

Structure-Based Drug Design

Generate 3D drug molecules inside protein binding pockets using decomposed diffusion. Produces molecules with drug-likeness scores, docking poses, and scaffold-arm decomposition.

Guan et al., ICML 2023 · CC-BY-NC 4.0
BoltzGen

Universal Binder Design

Design protein binders (nanobodies, peptides, mini-proteins) for any target using Boltzmann-distribution generative models. Produces diverse binder candidates with binding scores.

Stark et al., MIT 2025 · MIT License
ESMFold

Protein Structure Prediction

Predict 3D protein structures from amino acid sequences using the ESM-2 language model. Produces PDB structures with per-residue confidence scores (pLDDT) and 3D visualization.

Lin et al., Science 2023 · MIT License

Looking Forward

InsilicoΣ is more than a collection of tools, it is a growing ecosystem where research meets application. Our roadmap includes expanding deeper into bioinformatics and genomics, as well as new domains such as polymer informatics, multi-objective molecular optimization, and integrated hit-to-lead pipelines.

We are also planning to develop a user-friendly machine learning tool for clinical data analysis, bringing the same accessible, no-code approach to biomedical and clinical research. The goal is to empower clinicians and biomedical researchers to apply modern ML methods to patient data, biomarkers, and clinical outcomes without requiring programming expertise. Additionally, we aim to expand into genomics analysis tools, enabling researchers to perform gene expression analysis, pathway mapping, and variant interpretation through the same intuitive, no-code interface.

We are committed to keeping the platform free for academic researchers and continuously improving based on community feedback. Every tool we release carries a simple promise: if we publish a method, we will make it usable.

AI Lab